Hi there, I am analysing a highly zero-inflated dependent variable and using the code for a zero-inflated negative binomial (ZINB) model as per the user's guide: COUNT = u1- u4 (nbi); I have one main independent variable and a covariate (gender). Given that the ZINB output is in 2-parts (inflation and count), I wanted to generate a summary p-value & Chi2 statistic by using a likelihood ratio test to compare my main model against the null model (only gender predicting the dep. variable). I cannot work out how to do this, particularly as the "MODEL FIT INFORMATION" in the output does not give me a Chi2 statistic. Any help would be greatly appreciated! Thanks!
The chi-square that compares the observed to the model estimation covariance matrix is not available with count data because means, variances, and covariances are not sufficient statistics for model estimation. For two models with the same set of dependent variables, a difference test of two nested models can be carried out using the loglikelihoods. -2 times the loglikelihood difference is distributed as chi-square.